This article extends earlier research on the predictability of residential construction activity in regionalhousing markets. This category of forecasts is used in numerous banking, government, utility, andretail applications. To analyze forecast accuracy, quarterly frequency data are assembled from previouslypublished econometric forecasts for Florida and its six largest metropolitan economies. Thesample simulation covers the period between the first quarter of 1985 and the second quarter of 1996and includes all three business cycle phases: expansion, recession, and recovery.Forecasts for single-family starts are compared with univariate time series and random-walk alternatives.Results indicate that structural model forecasts of regional housing construction are comparativelyless reliable than forecasts for nonagricultural employment. Moreover, single-family starts inFlorida do not perform as well against the two benchmarks as multifamily counterparts do for the samehousing markets. Modelers should possibly consider increasing coverage to include other variables asa means of improving the reliability of residential construction forecasts.